和声搜索聚类优化模型的PPI功能模块挖掘算法研究
发布时间：2019-01-19 11:56
【摘要】：蛋白质交互(Protein-Protein Interaction,PPI)网络是生物体内蛋白质之间相互作用形成的网络,在拓扑结构上呈现小世界特性和无尺度特性,属于复杂网络的一种。近年来,随着高通量技术的发展,可获得的蛋白质交互数据日渐丰富,基于蛋白质交互网络的功能模块挖掘有助于预测未知蛋白质功能,为疾病研究提供理论基础,已成为生物信息学领域新的研究热点。与此同时,智能算法由于在解决复杂问题方面的优越性获得了广泛的应用,基于智能计算的算法被陆续应用在蛋白质交互网络的功能模块挖掘问题上,逐渐成为新的研究热点。本文将和声搜索算法应用在蛋白质交互网络的功能模块挖掘问题上并进行了深入的研究,主要工作包括:(1)基于和声搜索算法,提出了基于和声搜索(Harmony Search,HS)聚类优化模型的蛋白质交互网络功能模块挖掘算法(HMS-FMD),算法改进了传统和声搜索的搜索策略,在蛋白质交互网络中,将搜索聚集系数较大的结点集合作为算法的目标函数。通过实验对算法的参数进行分析和对比,得到了算法参数的最优设置,与其他挖掘算法相比,实验结果表明本文算法能有效挖掘出蛋白质交互网络中的功能模块。(2)当前的研究普遍将蛋白质网络看作一个边存在确定性的无向图,但由于高通量生物检测技术对蛋白质交存检测存在固有的误差,因此实验测得的蛋白质是否真实存在交互性是不确定的。在不确定图数据挖掘问题上,蛋白质功能模块挖掘问题的计算复杂性通常要比确定图数据同一挖掘问题的计算复杂性要高。本文利用“可能世界”模型,在不确定性蛋白质交互网络的基础上,提出了基于和声搜索优化模型的不确定蛋白质交互网络功能模块挖掘算法,通过理论推导,简化了计算复杂度,使用和声搜索聚类优化模型,将期望密度较大的结点集合作为算法搜索的目标函数。通过实验对算法进行分析和对比,结果表明该算法具有较好的聚类结果。本文通过对和声搜索聚类优化模型的算法研究,并应用在蛋白质交互网络功能模块挖掘问题上,在一定程度上丰富了蛋白质交互网络功能模块挖掘算法的理论研究,为蛋白质交互网络的功能模块挖掘研究提供一定的理论指导。
[Abstract]:Protein interaction (Protein-Protein Interaction,PPI) network is a network formed by the interaction of proteins in organisms. It has the characteristics of small world and no scale in topological structure. It belongs to one of the complex networks. In recent years, with the development of high-throughput technology, more and more protein interaction data are available. The function module mining based on protein interaction network is helpful to predict the unknown protein function and provide the theoretical basis for disease research. It has become a new research hotspot in the field of bioinformatics. At the same time, due to the advantages of intelligent algorithms in solving complex problems, intelligent computing algorithms have been gradually used in the protein interactive network functional module mining problem, gradually becoming a new research hotspot. In this paper, the harmonic search algorithm is applied to the functional module mining problem of protein interactive network. The main work is as follows: (1) based on the harmony search algorithm, the (Harmony Search, based on harmony search is proposed. HS) clustering optimization model of protein interactive network functional module mining algorithm (HMS-FMD), the algorithm improved the traditional harmony search strategy, in protein interactive networks, The set of nodes with large searching aggregation coefficient is regarded as the objective function of the algorithm. Through the analysis and comparison of the parameters of the algorithm, the optimal setting of the algorithm parameters is obtained, which is compared with other mining algorithms. Experimental results show that the proposed algorithm can effectively mine the functional modules in protein interaction networks. (2) the current research generally regards protein networks as an undirected graph with deterministic edge. However, due to the inherent error of high-throughput biological detection technology for protein deposit detection, it is uncertain whether the protein measured by experiments is truly interactive or not. In the problem of uncertain graph data mining, the computational complexity of protein functional module mining problem is usually higher than that of determining the same mining problem of graph data. In this paper, based on the uncertain protein interaction network and the "possible world" model, an algorithm for mining the functional modules of uncertain protein interaction networks based on harmony search optimization model is proposed. The computational complexity is simplified and the optimal model of harmonic search clustering is used. The set of nodes with high expected density is taken as the objective function of the algorithm. The experimental results show that the algorithm has better clustering results. In this paper, the algorithm of clustering optimization model of harmony search is studied, and it is applied to the problem of protein interactive network functional module mining. To some extent, it enriches the theoretical research of protein interactive network functional module mining algorithm. It provides some theoretical guidance for the research of functional module mining of protein interactive network.
【学位授予单位】：江西理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：Q51;TP311.13
[Abstract]:Protein interaction (Protein-Protein Interaction,PPI) network is a network formed by the interaction of proteins in organisms. It has the characteristics of small world and no scale in topological structure. It belongs to one of the complex networks. In recent years, with the development of high-throughput technology, more and more protein interaction data are available. The function module mining based on protein interaction network is helpful to predict the unknown protein function and provide the theoretical basis for disease research. It has become a new research hotspot in the field of bioinformatics. At the same time, due to the advantages of intelligent algorithms in solving complex problems, intelligent computing algorithms have been gradually used in the protein interactive network functional module mining problem, gradually becoming a new research hotspot. In this paper, the harmonic search algorithm is applied to the functional module mining problem of protein interactive network. The main work is as follows: (1) based on the harmony search algorithm, the (Harmony Search, based on harmony search is proposed. HS) clustering optimization model of protein interactive network functional module mining algorithm (HMS-FMD), the algorithm improved the traditional harmony search strategy, in protein interactive networks, The set of nodes with large searching aggregation coefficient is regarded as the objective function of the algorithm. Through the analysis and comparison of the parameters of the algorithm, the optimal setting of the algorithm parameters is obtained, which is compared with other mining algorithms. Experimental results show that the proposed algorithm can effectively mine the functional modules in protein interaction networks. (2) the current research generally regards protein networks as an undirected graph with deterministic edge. However, due to the inherent error of high-throughput biological detection technology for protein deposit detection, it is uncertain whether the protein measured by experiments is truly interactive or not. In the problem of uncertain graph data mining, the computational complexity of protein functional module mining problem is usually higher than that of determining the same mining problem of graph data. In this paper, based on the uncertain protein interaction network and the "possible world" model, an algorithm for mining the functional modules of uncertain protein interaction networks based on harmony search optimization model is proposed. The computational complexity is simplified and the optimal model of harmonic search clustering is used. The set of nodes with high expected density is taken as the objective function of the algorithm. The experimental results show that the algorithm has better clustering results. In this paper, the algorithm of clustering optimization model of harmony search is studied, and it is applied to the problem of protein interactive network functional module mining. To some extent, it enriches the theoretical research of protein interactive network functional module mining algorithm. It provides some theoretical guidance for the research of functional module mining of protein interactive network.
【学位授予单位】：江西理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：Q51;TP311.13
【参考文献】
相关期刊论文 前10条
1 郭琦;卢意力;李s，
本文编号：2411356
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2411356.html
